Serialbench

Measurement console · run 2026-08-24

How fast can Ruby read and write your data?

Every XML, JSON, YAML and TOML serializer in the ecosystem, measured on the same workloads across34platform and Ruby combinations. Pick a format, pin a reference, and read the field.

Format
Operation

Field ranking — iterations per second, higher is better

macOS 14 · arm64 · Ruby 3.2 · Parse · large input

0.000.000.000.000.00
vs ref
Cite this
“Pin a reference library to generate a citable claim.”

Will it run on my stack — coverage across measured environments

Derived from the runs themselves: a dot means the library loaded and completed the benchmark on that platform.

LibrarymacOS 14arm64 · 3.2macOS 14arm64 · 3.3macOS 14arm64 · 3.4macOS 15 Intelx64 · 3.2macOS 15 Intelx64 · 3.3macOS 15 Intelx64 · 3.4macOS 15arm64 · 3.2macOS 15arm64 · 3.3macOS 15arm64 · 3.4macOS 26 Intelx64 · 3.2macOS 26 Intelx64 · 3.3macOS 26 Intelx64 · 3.4macOS 26arm64 · 3.2macOS 26arm64 · 3.3macOS 26arm64 · 3.4Ubuntu 22.04arm64 · 3.2Ubuntu 22.04arm64 · 3.3Ubuntu 22.04arm64 · 3.4Ubuntu 22.04x64 · 3.2Ubuntu 22.04x64 · 3.3Ubuntu 22.04x64 · 3.4Ubuntu 24.04arm64 · 3.2Ubuntu 24.04arm64 · 3.3Ubuntu 24.04arm64 · 3.4Ubuntu 24.04x64 · 3.2Ubuntu 24.04x64 · 3.3Ubuntu 24.04x64 · 3.4Windows 11arm64 · 3.4Windows 2022x64 · 3.2Windows 2022x64 · 3.3Windows 2022x64 · 3.4Windows 2025x64 · 3.2Windows 2025x64 · 3.3Windows 2025x64 · 3.4
json
leptris
nokogiri
oga
oj
ox
psych
rapidjson
rexml
toml-rb
tomlib
tomlrb
yajl

Raw data

Full results per environment and the complete resultset are exported as YAML alongside this site —data/<environment>.yamlanddata/resultset.yaml. Re-run any leg withserialbench environment executeusing the configs in the repository.